A/B tests need to be as specific to one adjustment as possible.
I've noticed flaws like this many times on ABtests.com. But I guess they are like any war stories, its the overall idea that counts not the details.
What are the primary flaws you're seeing?
In your case adding a headline to the form was overlooked.
You were using form submissions as a metric from what I can tell. So changes specifically to the signup form itself (such as the header) is likely to have as much of an impact as the video.
Did you track how many people watched the video then signed up as a metric? That may be more telling.